A Hybrid Deep Learning and Residual Kriging Method for Private 5G Radio Map Generation

As e-Um 5G (private 5G) has emerged as a key infrastructure for digital transformation, accurate high-resolution radiowave mapping has become crucial for stable service and efficient deployment. However, traditional kriging has a limited ability to model the complex nonlinearities of radio environments. To enhance e-Um 5G map accuracy, this study proposes a novel hybrid interpolation method that combines deep learning and residual kriging. Using actual data, the proposed method was found to be superior to traditional kriging in terms of the prediction accuracy and map resolution, particularly with varying sample sizes.